TL;DR
This paper introduces ORNARE, an integrated digital tool designed to create a comprehensive onomastic repertoire for medieval French romance. By combining advanced digital humanities techniques with traditional philological methods, ORNARE aims to unlock deeper insights into the cultural and linguistic nuances embedded within these historical texts.
What Problem Does This Paper Address?
Medieval French romances, particularly those centered around the figure of Alexander the Great (the Roman d'Alexandre or RdA), are rich with proper names that offer crucial insights into cultural norms, literary conventions, and linguistic evolution. However, studying these names presents significant challenges:
- Onomastic Variability: Medieval texts exhibit extreme variability in spelling, abbreviations, and forms of proper names, making consistent identification and normalization difficult.
- Ambiguity of "Proper Name": Distinguishing true proper names from common nouns used in specific contexts (e.g., "Riche Chevalier") or epithets is a complex philological task.
- Genre Definition: The very definition of "romance" in the Middle Ages is fluid, complicating corpus selection.
- Lack of Digital Tools: Existing onomastic tools for medieval literature are often outdated, lack comprehensive corpora, or are not designed for collaborative, in-depth analysis.
This paper addresses the need for a modern, integrated digital solution that can handle this complexity, facilitate collaborative research, and make medieval onomastic data more accessible and analyzable.
Methods
The authors propose a step-by-step methodology for designing and developing a web application for onomastic annotation and analysis, focusing on an integrated environment for linguists and philologists. Key methodological steps include:
- OCR and Data Extraction: Utilizing tools like Tesseract to convert scanned editions of medieval romances into machine-readable text.
- Cleaning and Pre-processing: Removing extraneous characters and formatting to prepare the text for analysis.
- Data Structuring and Normalization: Organizing text, identifying and standardizing name variants (including misspellings and abbreviations), and linking names to their contextual information.
- Data Annotation: Enabling manual verification and annotation by experts, establishing guidelines for consistency, and validating data quality.
- Named Entity Recognition (NER): Employing NLP techniques to automatically recognize and extract named entities.
- Data Curation and Harmonization: Resolving discrepancies between manual and automated extraction, and standardizing name representations.
- Collaboration and Data Sharing: Implementing features for multiple researchers to work simultaneously and share data.
The ORNARE system itself is built using the R Shiny framework, incorporating the datatable (DT) package for annotation review and the textplot package for text visualization. The system allows users to search by character, normalized name, or specific spelling, and to directly edit verses.
Datasets
The primary dataset for this project is the Medieval French Roman d'Alexandre (RdA) corpus. This corpus consists of approximately 1500 pages from critical editions, containing over 33,000 verses. The authors note that while other digital corpora for Old French exist, the RdA was chosen for its significance and the lack of a comprehensive Index nominum in existing editions. The project aims to create a new digital onomastic repertoire for this corpus.
Key Insights and Ideas
- The Power of Integrated Digital Tools: The paper highlights how combining OCR, NLP, and interactive web interfaces can overcome the significant challenges of medieval onomastic research, which has traditionally been labor-intensive and prone to inconsistencies.
- Onomastics as a Window into Medieval Culture: The variability and specific choices in naming within medieval romances are not random but reflect cultural modalization, stereotypes, and evolving linguistic practices. ORNARE aims to make these patterns discoverable.
- Modular and Integrable Design: The ORNARE system is designed to be modular, allowing for the addition of new romances and names, thus broadening its scope and utility for future research.
- Bridging Philology and Digital Humanities: The project emphasizes a collaborative approach, where philologists' expertise in textual interpretation is enhanced by digital tools, and computer scientists benefit from the unique challenges presented by historical texts.
- OntoLex-Lemon and TEI for Interoperability: The authors plan to publish linguistic data using the OntoLex-Lemon paradigm and TEI XML format to ensure interoperability and semantic richness within the broader linguistic and semantic web community.
Key Results
- The project has successfully designed and implemented the ORNARE system, an integrated web environment for the analysis of medieval French romance onomastics.
- Initial work has identified 2,219 spellings of names within the RdA corpus. The exact number of distinct characters and normalized names is still under evaluation due to philological complexities.
- The system allows for dual search functionalities: by specific name spellings or by character types/romances.
- An experimental test involved semi-automatic filtering of capitalized words at the beginning of verses, though manual skipping was preferred for initial annotation to avoid losing valuable data.
Limitations and Future Work
- OCR Accuracy and Correction: The paper acknowledges that the raw OCR text was used without modification in the initial phase to avoid introducing unwanted corrections. The semi-automatic correction of OCRed text is planned for future work.
- Database Design: The design and implementation of the relational database to store data and annotations are also noted as parts of the implementation intentionally left out for this paper.
- Scope Expansion: Future work will involve expanding the corpus to include more romances and names, further refining the annotation process, and developing more sophisticated filtering and analysis capabilities.
- Integration with CLARIN: Integrating the annotated dataset into the CLARIN framework is a strategic step for long-term preservation and accessibility.
- Full Publication of Implementation Details: The authors intend to present the OCR process, semi-automatic correction, and database design in a future, extended version of the paper.
Summary updated